An important cause of instability in the balance of payments of primary producing countries has been instability in the international prices of primary commodities. Price instability can also cause inflation in the countries that import primary commodities—many of which are industrial countries.1
Perhaps because both consuming and producing countries consider price instability undesirable, the drastic changes in prices of primary commodities between 1973 and 1975 attracted much attention and caused particular concern. 2 It is against this background that the United Nations Conference on Trade and Development (UNCTAD) has recently proposed an Integrated Program on Commodities (See UNCTAD (1974).) to stabilize international prices of primary commodities.
The fact that the fluctuations in international prices of primary commodities have been much more pronounced than those in international prices of manufactured goods has often been rationalized on the grounds that international commodity markets are more competitive than the domestic markets for manufactures. This is so because prices are more responsive to shifts in supply and demand in competitive markets than they are in monopolistic or oligopolistic markets. In fact, in the latter, theoretical and empirical evidence has shown that prices are, more often than not, determined by factors that bear little, if any, relationship to short-run variations in demand and supply.
While it has been a well-known proposition even in elementary textbooks of economics that prices in competitive markets respond to demand and supply, economists, as noted by McCallum (1974), have made very little effort to formulate an “operational” version of this proposition so that it could be subjected to an econometric test. McCallum’s attempt to formulate and test a competitive model of price adjustment using data for the U. S. lumber industry is thus a notable exception. In contrast, there have been numerous studies concerned with the formulation and testing of monopolistic models of price dynamics.
In this paper, I will formulate a dynamic disequilibrium model of price adjustments in competitive markets and will subsequently subject this model to an empirical test using the data for seven primary commodities in the UNCTAD proposal. 3 While my attempt here is similar to the McCallum study in spirit, I will formulate and test a competitive model of price adjustment in a structural form. 4 By contrast, the McCallum model was formulated and tested in a reduced form. In addition to being more explicit, a structural model has the advantage over a reduced-form model, in that the former yields structural information about the model that is often buried in the latter.
The purposes of this paper are several. First, it provides a test as to whether the marked fluctuations in the prices of primary commodities traded in the international markets could be well explained by the systematic variations of demand and supply and provides a test of how a competitive model of price adjustments really fares when used to explain the price behavior of primary commodities. Second, it throws light on the price determination mechanism for primary commodities in general, because, as shall be argued later, the empirical literature on competitive price models applied to the international markets of primary commodities has been ambiguous. In particular, the literature has failed to indicate whether international commodity prices are primarily stock-, or flow-, or mixed stock-flow-determined. Finally, it facilitates the analytical assessment regarding the feasibility of stabilizing primary commodity prices through buffer stocks, as is envisaged in the UNCTAD proposal, and/or through other means. For there is no doubt that economic analysis of the likely effects of buffer stocks on commodity prices and export earnings has to rely upon knowledge of the price dynamics in international commodity markets.
The plan of this paper is as follows: Section I gives a brief review of the competitive, structural models of price formation that have been used to analyze international commodity prices. Section II formulates a competitive “stock” disequilibrium model of price adjustments. Section III sets forth the empirical tests of this model, and Section IV gives a summary of conclusions.
Adams, F. G., and Jere R. Behrman, Econometric Models of World Agricultural Commodity Markets: Cocoa, Coffee, Tea, Wool, Cotton, Sugar, Wheat, Rice (Cambridge, Massachusetts, 1976).
McCallum, B. T., “Competitive Price Adjustments: An Empirical Study,” American Economic Review, Vol. 64 (March 1974), pp. 56–65.
Telser, Lester G., “Futures Trading and the Storage of Cotton and Wheat,” Journal of Political Economy, Vol. 66 (June 1958), pp. 233–55.
United Nations Conference on Trade and Development, An Integrated Programme for Commodities, U. N. Document No. TD/B/C. 1/166 (United Nations, Geneva, 1974).
Mr. Hwa, economist in the Commodities Division of the Research Department, holds degrees from Cornell University. He has held research positions at Cornell University’s Center for Quantitative and Mathematical Research in Economics and Management, at the National Bureau of Economic Research, and at the Stanford Research Institute.
In addition to colleagues in the Fund, the author would like to thank P. Schelde Anderson, Sarra Chernick, Gary Fromm, David Crary, and Walter Labys for their very helpful comments.
It is sometimes argued that boom-and-bust cycles in primary commodity prices cannot be the underlying cause of general inflation, for they can affect only relative prices—not all money prices. But there are situations in which these cycles can have inflationary consequences, such as when the ratchet effect is at work.
The aggregate spot export price index (1970=100) of primary commodities compiled by the International Monetary Fund reached 212.1 in 1974—its highest level since 1957—and then dropped to 174.1 in 1975 (See the Fund’s monthly publication, International Financial Statistics.). The annual percentage changes in this index during 1973, 1974, and 1975 were, respectively, 54.7, 27.8, and -17.9. The aggregate index masks the fluctuations in its constituent commodity prices, which often display much greater variation.
The seven commodities are cocoa, coffee, sugar, rubber, cotton, tin, and copper.
The model is structural in the sense that, as will be clear later, consumption and production are not substituted for their respective underlying determinants.
See Walter C. Labys (1973, pp. 85–105). In this study, I will only be concerned with storable commodities.
Barro (1972) has taken a different view. He argues that the response of prices to disequilibrium is essentially a monopolistic phenomenon.
If flow disequilibrium is defined as the gap between imports and exports, rather than between consumption and production, flow disequilibrium is equivalent to stock disequilibrium. The proof is given here. I write two equations defining, respectively, aggregate exports and aggregate imports of a commodity. And, without loss of generality, I assume that exports and imports are, respectively, net exports and net imports. Thus
See Working (1949), Telser (1958), and Brennan (1958). Meir Kohn (1978) has shown that such a relationship can also be derived from the rational behavior of individual speculators. Thus,
It seems that a natural candidate for
For instance, there is no reason for the price predictions generated by a distributed-lag hypothesis to be consistent with those implied by the underlying model. However, the argument is not meant to imply that the rational expectations hypothesis is free from any defect. Indeed, the rational expectations hypothesis might give market participants more information than they actually possess.
The world consumption function of a commodity is postulated to be a function of its own price and world money income,
Similarly, the level of output that maximizes a short-run profit function, subject to the constraint of a short-run production function under perfectly competitive conditions of both product and factor markets, can be shown to be
Equations (7), (15), (17), and (18) form a simultaneous-equation model that determines four endogenous variables—Ct, Qt, Pt, and Ht. The predetermined variables are PMt, Yt, T, and Pt–1. However, the time trend is excluded from the list of predetermined variables in equation (16′) because it is highly correlated with the real world income variable.
In the equations for cotton and copper, PMt becomes highly significant after Yt and T are deleted.
Using a 5 per cent confidence interval, the Durbin-Watson statistics reveal that the hypothesis of first-order serial correlation should be rejected for all commodities, except for cocoa.
This simplification will reduce somewhat the predictive power that accrues to the market participants, but the reduction will probably not be considerable. It will affect the overall predictive power of the price equation even less, because Yt is likely to be collinear with both Ct and PMt, and Pt–1 is already in the price equation.
Private stocks are stocks held by private traders; they exclude those stocks held by national authorities and international buffer stock agencies.
The other three core commodities are jute, sisal, and tea.
The instruments used for each commodity are world income, the world price, and lagged commodity prices.
The variables with incorrect signs were deleted, and the regression was rerun. However, the variables that had correct signs were retained in the regression, even if they were statistically insignificant.
Note that the equations for cotton and sugar were corrected for first-order serial correlation, so that the lagged dependent variable is still present in each equation.
Because each version of the model has similar standard errors, only the results for the first version—OLS, equation (20′)—are reported here.
In the case of tin, the level of inventory is significant at 5 per cent for the INV estimation of equation (20″) (See the equation for tin in Table 5.), although it is not as significant in the other three estimations.
This is not to say that the flow variables are not important in determining price; on the contrary, they are important.
The constrained version—that is, equation (20″)—has smaller standard errors for cocoa, coffee, and sugar in the OLS results; and, in addition to these three commodities, it also has smaller standard errors for cotton in the INV results.
The elasticities e are evaluated at sample means—that is, e =
The coefficient of Pt–1 is not significantly different from zero, which implies that the speed of price adjustment r approaches infinity.
The tin equation was excluded from this exercise because a structural shift that occurred in 1974 was detected during the process of estimation.